Season Predictions

How many games will a college football team win in the upcoming season? How likely are they to win their conference championship?

This notebook displays 2022 predictions (as of Week 2) and historical data for SMU.

Note: a team’s predicted win/loss total is the average number of wins they achieve across all simulations of games in the regular season.

2022 Game Predictions

Using each team’s rating, I simulate the result and margin of victory for each remaining game of the 2022 season 1000 times.

The following table shows the percentage of times in which SMU won each of the games on their upcoming schedule. The margin of victory is the median margin of victory across all simulations.

Season Week Date Team Opponent Pr(Win) Prediction Result Correct?
2022 1 Sep 03 @ 0.565 win by 3 won by 38 ✓
2022 2 Sep 10 vs 0.893 win by 22 won by 29 ✓
2022 3 Sep 17 @ 0.429 lose by 3
2022 4 Sep 24 vs 0.642 win by 6
2022 5 Oct 01 @ 0.486 lose by 1
2022 7 Oct 14 vs 0.909 win by 23
2022 8 Oct 22 vs 0.452 lose by 2
2022 9 Oct 29 @ 0.568 win by 3.5
2022 10 Nov 05 vs 0.588 win by 4
2022 11 Nov 12 @ 0.744 win by 15
2022 12 Nov 17 @ 0.597 win by 6
2022 13 Nov 26 vs 0.692 win by 9

It’s important to note that within each simulation the results of each game are not independent. The (simulated) results of each week are used to update each team’s rating going into the next week, which will then influence how likely they are to win games in the rest of the season. This means that in some simulations a team will overperform/underperform early and end up with a very good/poor season. A team’s predicted win total is the average number of wins they achieve across all simulations.

Predicted Win Totals

The visualization below shows how SMU’s predicted win total has changed after each week of the regular season so far.

Simulated Rating for Rest of Season

The table above indicates the aggregated result of every simulated game for the selected team. The visualization below shows the simulated paths for SMU as of Week 2 in the 2022 season. This illustrates the most common paths left in a team’s season by showing how the result of one game tends to affect its future trajectory.

Rating vs Preseason Expectations

How is SMU performing compared to preseason expectations? The visualization below shows the team’s rating so far compared to simulated ratings for the team from the start of the season. The simulated paths from the beginning of the season are shown in grey while the team’s actual rating so far is shown in the team’s color. Teams that are outperforming preseason expectations will be above the majority of the simulated paths; teams that are under performing will be below.

Conference and Playoff Probabilities

How has SMU’s chance’s of winning a conference championship and making/winning the playoffs changed over the course of the season?

Historical Team Data

The simulations for an upcoming season make use of an (adjusted) Elo rating, which is a rating assigned to each team based on wins and losses in previous games. After wins, a team’s rating will go up; after losses, a team’s rating will go down. The magnitude of these increases/decreases will vary depend on the scoreline and strength of the opponent.

Elo Rating

The following visualizations shows SMU’s Elo rating from 1900 to 2022

Efficiency

In addition to a team’s historical Elo rating, I also compute and use each team’s offensive/defensive efficiency metrics from recent seasons to predict their expected future Elo rating. These efficiency metrics come from an expected points model I trained on on play by play data in order to identify the value of individual plays within a game.

I score every play with the model and then aggregate these results to the game and season level and adjust for opponent quality. These results are only available starting from the season of 2007, as this is when reliable play by play data becomes available.

For more details on what goes into creating these ratings, go to my description of the expected points model and my methodology for adjusting for opponent quality.

These efficiency ratings indicate a team’s expected points per play when its offense or defense is on the field, adjusted for opponent quality. A team’s overall rating is a combination of its offense and defense ratings, and indicates the net points per play a team would expect when playing an average opponent.

For offenses, this indicates the average points the team scores against opponents per play. For defenses, this indicates the average points the team prevents opposing offenses from scoring per play. In both cases, I have set the scale of the variable to mean that positive is good while negative is bad.

The visualization below shows how a team has changed over time as well as their season end efficiency ranking (out of all FBS teams).

Efficiency by Play Type

In addition to examinig how a team performs overall, we can examine each team’s efficiency based on the play type. How has SMU performed when running/passing on offense vs defending the run/pass on defense?

The following visualization shows how efficient SMU’s offense has been running/passing the ball each season since 2007.

The following visualization shows how efficient SMU’s defense has been in stopping the opponent’s run/pass in each season since 2007.